CN1387722A - Image compressing method - Google Patents
Image compressing method Download PDFInfo
- Publication number
- CN1387722A CN1387722A CN00814924A CN00814924A CN1387722A CN 1387722 A CN1387722 A CN 1387722A CN 00814924 A CN00814924 A CN 00814924A CN 00814924 A CN00814924 A CN 00814924A CN 1387722 A CN1387722 A CN 1387722A
- Authority
- CN
- China
- Prior art keywords
- data
- mentioned
- compressing image
- view data
- image data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/41—Bandwidth or redundancy reduction
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/46—Colour picture communication systems
- H04N1/64—Systems for the transmission or the storage of the colour picture signal; Details therefor, e.g. coding or decoding means therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
- G06T9/005—Statistical coding, e.g. Huffman, run length coding
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Compression Of Band Width Or Redundancy In Fax (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
- Image Processing (AREA)
- Color Television Systems (AREA)
Abstract
Bit-map digital image data with one or more color channels is divided into optimum processing units of image size preset in response to a requirement for the throughput of the computer and the quality of image, the image data units are further divided into minimum processing units of image data for each color channel. The minimum processing units of image data of each column in the lateral direction are converted into two-dimensional vectors with the lateral position x and the luminance y of each pixel and quantized according to the magnitude of the change of the luminance by using the line segment connecting the start and end points of each column as a reference vector and shaping is carried out. Then the minimum processing units of image data of each line in the vertical direction are similarly converted into two-dimensional vectors with the vertical position z and the luminance y and quantized according to the magnitude of the change of the luminance by using the line segment connecting the start and end points of each line as a reference vector. Thus the data is converted into compressed image data of a matrix structure based on the significance of the luminance information that each pixel has. The compressed image data of minimum processing units are arithmetically compressed, and then compressed image data of optimum processing units is created by integrating all the color channels. By further integrating them, compressed image data on the whole original image is created.
Description
Technical field
The present invention relates to compress for example image processing techniques of the DID of the full-color image of additive color three primary colors (RGB) or the three primary colors of losing lustre (CMY) etc. or 256 gray level images of black and white etc.
Background technology
In general, in the Computer Processing of coloured image, used by each colouring component in the color model of RGB system or CMY system etc., be the DID of the monochrome information formation of each chrominance channel (color model).This DID for example distributes in every kind of color of RGB the information of 1 byte (8) that the value (256 gray scale) with 0～255 256 grades shows to 1 pixel, thereby has the information of 3 bytes in each pixel.Therefore, the data volume of particularly high meticulous coloured image is huge, for such data volume is carried out Computer Processing, not only need the hardware resource of jumbo memory or hard disk and high-speed CPU etc. and can carry out the communication technology that data at a high speed transmit, and need the technology of compressing image data.
Propose various data compression methods in the past, but when restoring the image that has compressed, produced the situation of quality variation sometimes.For example, because view data is not necessarily used whole colors, so the method that does not influence the original image of image by carrying out the losing lustre the size that reduces file of appropriateness is in advance arranged.In following such compression method of losing lustre, owing in compression process, merge approximate look information,, particularly there is the reproduction of the fine rule feint trend that becomes so lost the continuity of tone variations.In addition, known have monochrome information is transformed to the compression method that colour difference information keeps look information.But the amount of information of the number of significant digit of colour difference information is the about below 1/4 of monochrome information, has the trend of the deflection that causes tone in the reproduction process.
In addition, open in the flat 10-32719 communique the spy, high reproduction quality with sufficient compression ratio and image is a purpose, discloses and will be transformed to the color model data that are made of 1 monochrome information and 2 colour difference informations by the color image data that trichromatic monochrome information constitutes, with reproducing the method for compressing image and the device of the method compression of these data with high-grade.As this compression treatment technology that can reproduce with high-grade, adopted so-called trivector to quantize, promptly, in the XZ plane of the Luminance Distribution of representing each pixel in the XY plane of delineation with the size of Z direction, come apparent brightness can think the pixel groups of 1 group of horizontal direction in the scope of certain admissible error, changing continuously with the vector of 1 unit, and, equally in the YZ plane, come apparent brightness can think the pixel groups of 1 group of vertical direction in the scope of certain admissible error, changing continuously with the vector of 1 unit.In the method for compressing image of this communique, even compared with the pastly can improve compression ratio significantly, but owing to be transformed to colour difference information and compress, so when making it to restore, have the possibility of reproduced image deterioration in that view data is decompressed.
Therefore, the object of the present invention is to provide a kind of like this method for compressing image that reproduces high quality images, wherein, even for the meticulous DID of height, also both can realize high compression effectiveness, can reproduce the continuous variation of tone again to heavens, and this moment even can reproduce fine rule brightly, the use of picture original can be stood thus as printing.
Disclosure of an invention
According to the present invention, following method for compressing image is provided, it is characterized in that:
Each row for the transverse direction of the bitmap with single or multiple chrominance channel, DID, position and brightness with the transverse direction of each pixel are carried out 2 n dimensional vector nizations to this view data, and be benchmark with the part of line of the Origin And Destination that links above-mentioned each row, the size of utilizing brightness to change quantizes
Each row for the longitudinal direction of above-mentioned view data, position and brightness with longitudinal direction are carried out 2 n dimensional vector nizations to the above-mentioned view data of having carried out vector quantization and quantification on transverse direction, and be benchmark with the part of line of the Origin And Destination that links above-mentioned each row, the size of utilizing brightness to change quantizes, thus, be transformed to the compressing image data of matrix structure of the number of significant digit of the monochrome information that has based on each pixel.
Like this, owing to quantize by the size that the monochrome information of each pixel of comprising in the view data is carried out 2 n dimensional vector nizations and utilized brightness to change, the degree of the number of significant digit of the information that has according to each pixel is carried out stratification to data and is handled once more and constitute, so will there not be the monochrome information of the pixel of number of significant digit to absorb in the identical vector in fact.Thereby, can reduce its data volume and do not lose the information that comprises in the original view data in fact, can be made into and be particularly suitable for data structure high meticulous view data, that compression efficiency is the highest.
In certain embodiments, original bitmap, DID are divided into the unit view data of given size, for the constituent parts view data of having cut apart, carry out the vector quantization and the quantification of transverse direction respectively, and carry out the vector quantization and the quantification of longitudinal direction, be transformed to the unit compressing image data of matrix structure, can merge the constituent parts compressing image data at last.Thus, even under the bigger situation of the data volume of original painting, also can handle expeditiously according to the disposal ability or the desired picture quality of computer.
In a further embodiment, original bitmap, DID are divided into each chrominance channel, after the view data of each chrominance channel that will cut apart is transformed to the compressing image data of matrix structure respectively, can be merged into 1 compressing image data.Thus, owing to for example view data for RGB system or CMY system can be reduced to 1/3 with handling unit, so can handle expeditiously.
In addition, under the situation of the unit view data that original bitmap DID is divided into given size, after the unit view data of each chrominance channel that this constituent parts view data is divided into each chrominance channel again, will have cut apart is transformed to the unit compressing image data of matrix structure respectively, by being merged into 1 unit compressing image data, can carrying out the higher compression of efficient and handle.
Moreover, according to the present invention, can be again the compressing image data of the matrix structure that generates be by this way carried out the arithmetic compression.Thus, mode that can non-loss is compressing image data more to heavens.
In certain embodiments, the process of the quantification of transverse direction and longitudinal direction is made of following process, and promptly the part with the line of the Origin And Destination that links each row or each row is a base vector, calculate in this interval just and/or negative maximum deviation point, in starting point and terminal point and maximum deviation point, part with each line of 2 of linking adjacency is new base vector, calculate the process of this maximum deviation point and repeat deviation about the base vector of each maximum deviation point till the number of significant digit of losing as desirable view data, utilize the size of the deviation of each maximum deviation point to distinguish each pixel, generate the process of the different many levels data of figure place respectively.
For example, original be under 8 the data conditions, the most handy the 1st～the 4th hierarchical data that is made of 8,6,4,1 data respectively constitutes above-mentioned many levels data.
The simple declaration of accompanying drawing
Fig. 1 is the flow chart that the preferred embodiment of method for compressing image of the present invention roughly is shown.
Fig. 2 illustrates the flow chart of process of step 3 that view data is divided into Fig. 1 of optimization process unit.
Fig. 3 roughly illustrates the view data that is divided into optimization process unit is carried out the flow chart of vector quantization with the step 4 of Fig. 1 of the view data of generator matrix structure.
Fig. 4 is illustrated on the transverse direction view data is carried out the flow chart of vector quantization with the process that generates the 1st hierarchical data.
Fig. 5 is illustrated on the transverse direction view data is carried out the flow chart of vector quantization with the process that generates the 2nd hierarchical data.
Fig. 6 is illustrated on the transverse direction view data is carried out the flow chart of vector quantization with the process that generates the 3rd hierarchical data.
Fig. 7 is illustrated on the transverse direction view data is carried out the flow chart of vector quantization with the process that generates the 4th hierarchical data.
Fig. 8 roughly is illustrated in the view data of having carried out vector quantization on the transverse direction is carried out the flow chart of shaping processing back in the process of the enterprising row vectorization of longitudinal direction.
Fig. 9 illustrates the view data of having carried out vector quantization on transverse direction and longitudinal direction is carried out the flow chart of reprocessing with the process that generates VFZ image file of the present invention.
Figure 10 is a flow chart of playing the process of the VFZ image file of the present invention that has generated.
Figure 11 is the figure that an example of cutting apart of view data is shown.
Figure 12 is the figure that the view data that will cut apart in Figure 11 is divided into each chrominance channel of R, G, B again.
Figure 13 A is the figure of view data that transverse direction 1 row of chrominance channel R are shown, Figure 13 B be Luminance Distribution is shown, promptly for the size of the brightness of locations of pixels (length of counting from starting point), be the line chart that the notion of this vector quantization of explanation is used.
Figure 14 A illustrates the figure that utilizes the 1st～the 4th hierarchical data that vector quantization generates by the view data of transverse direction 1 row, and Figure 14 B illustrates the figure that the view data of whole row of having carried out vector quantization on transverse direction has been carried out the state that shaping handles.
Figure 15 is the concept map of distribution that is illustrated in the 1st～the 4th hierarchical data of the view data of having carried out vector quantization on transverse direction and the longitudinal direction.
Figure 16 is the figure that is illustrated in the arrangement of the view data of having carried out vector quantization on transverse direction and the longitudinal direction.
The optimal morphology of the usefulness that carries out an invention
Below, Yi Bian with reference to accompanying drawing, Yi Bian use its preferred embodiment to explain method for compressing image of the present invention.
Fig. 1 is to use the method for embodiments of the invention that original copy from full-color image roughly is shown, is the flow chart that original painting generates the process of compressed image file.At first, desirable picture original is carried out electronics processing, generate view data (step S1) as the bitmap of 1 chrominance channel that has the such a plurality of chrominance channels of RGB system for example or CMY system or only show with gray scale.Use various known devices such as image analyzer, digital camera, to import resolution arbitrarily, to generate common such view data with the brightness of the amount of information of for example 12,8 (1 bytes) that preestablished, 1 (2 value) etc.In order to carry out image Compression of the present invention, through the recording medium of CR-R, MO dish, DVD etc. or used the network of Ethernet etc., view data is input to (step S2) in process computer or the work station with online mode.
Above-mentioned process computer carries out pre-treatment to the view data that has been transfused to, is stored in (step S3) in the memory.At first, as shown in Figure 2, the integral body of unfolded image data (step S21) on memory.Secondly, the information that comprises in the heading label to view data is analyzed, in the structure of the size of having confirmed integral image (counting), chrominance channel, distribute to the monochrome information amount (figure place) etc. of 1 pixel after, be divided into the picture size of the optimization process unit that is predetermined according to analysis result, in memory, constitute the constituent parts view data of having cut apart (step S22) once more.This optimization process unit can be redefined for certain picture size, perhaps, can suitably select this optimization process unit according to such image broadcast condition such as ability, the desired image quality of the processing speed of process computer or memory span etc. etc.
In the present embodiment, as shown in Figure 11,1024 * 1024 RGB color image datas are compressed processing.If optimization process unit is decided to be 256 * 256 points, then the view data of Figure 11 is cut apart by 4 respectively on longitudinal direction and transverse direction, constitutes whole 16 unit view data once more.Each pixel of this RGB unit's view data has 1 byte respectively, adds up to the amount of information (brightness value) with 3 bytes for each chrominance channel of R, G, B.Secondly, as shown in Figure 12, this unit view data is divided into the unit view data of each chrominance channel of R, G, B, on memory, constitutes (step S23) once more.The unit view data of this each chrominance channel as minimum treat unit, is carried out the compression of data and handled.Each pixel of the unit view data of single chrominance channel has the amount of information (brightness value) of 1 byte, thereby the data volume of minimum treat unit is from 1024 of the situation of handling original view data integral body
2* 3
2Byte is reduced to 256 * 256 bytes.
Secondly,, in each single chrominance channel, successively this unit view data is carried out resolution of vectors respectively and handle, generate the view data (step S4) that is transformed to matrix structure for each optimization process unit.As shown in Figure 3, in each chrominance channel, on transverse direction, in order the view data of unit view data is carried out vector quantization, the 1st～the 4th hierarchical data (step S31) that generation is distinguished by the size (figure place) of the monochrome information of each pixel by row.
Use Fig. 4 to Fig. 7 to explain the vector quantization of transverse direction.As shown in Figure 13 A, the view data of minimum treat unit is made of 256 continuous pixels of respectively being listed as of transverse direction, and each pixel has the brightness value by the amount of information performance of 1 byte respectively.For 1 row of transverse direction,, show Luminance Distribution among Figure 13 B about horizontal 1 row by getting each locations of pixels in the length till with 0～255 on the transverse axis, getting the brightness value of each pixel in 256 grades till with identical 0～255 on the longitudinal axis.In the present invention, show the monochrome information of each pixel, carry out the vector quantization of view data in the mode of the information of the continuous variation that keeps Luminance Distribution and brightness by it being replaced into 2 n dimensional vector ns with length x and brightness value y.
Specifically, with the 1st row initial pixel be starting point P1 (x1=0, y1), and with last pixel be terminal point P2 (x2=255 y2), is master vector P1P2=[x2-x1 with the part that links this line of 2, y2-y1] (step S41).Secondly, about other pixel in the interval that links Origin And Destination, be the deviation that benchmark calculates brightness with the master vector, decision get the maximum deviation of positive and negative some D1 (x11, y11), D2 (x12, y12) (step S42).And, between each 2 point, P1 and D1, D1 and D2 and the D2 and the P2 of adjacency, similarly generate P1D1=[x11-x1, y11-y1 respectively], D1D2=[x12-x11, y12-y11] and D2P2=[x2-x12, y2-y12] (step S43).In addition, though consider that deviation is benchmark with the master vector, the point of only a side of plus or minus, promptly getting maximum deviation is 1 situation, this moment also between each 2 of adjacency similarly from linking the part generation sub-vector of this line of 2.
At this moment, the absolute value of the deviation among decision maximum deviation point D1 and the D2 whether respectively below 64, be whether its amount of information at (step S44) below 6.Be not below 64, promptly than it under the big situation, the variable quantity for the brightness of starting point and terminal point that is evaluated as this D1 or D2 has should be with 8 amount of information that show.On the contrary, under the situation below 64, because being evaluated as that the variable quantity for the brightness of starting point and terminal point of this D1 or D2 has should be with 6 or its amount of information with the performance of getting off, so the generative process of the 1st hierarchical data in the interval of starting point and terminal point finishes, transfer to the generative process of the 2nd following hierarchical data.
The absolute value of maximum deviation than 64 big situations under, the above-mentioned sub-vector that has generated between each 2 of adjacency that will comprise this D1 and/or D2 is stored in (step S45) in the memory as the 1st hierarchical data.X, y2 the amount that will have 8 amount of information respectively as shown in Figure 14 A, arranged these sub-vectors as 1 data continuously.
Moreover, in the interval separately of each 2 point, P1 and D1, D1 and D2 and D2 and the P2 that link adjacency, be benchmark with each sub-vector, calculate the deviation of brightness, some D11, D12, D21, D22, D31, the D32 of the maximum deviation of positive and negative got in decision.Secondly, for whole maximum deviation point of these new decisions, each the step S43～S45 that repeats to carry out for initial maximum deviation point D1, D2 is below 64 up to the absolute value of these maximum deviations.
Promptly, as among each 2 of the sub-vector of the benchmark of these maximum deviation points, between each 2 of adjacency respectively, similarly generate (the step S43) such as sub-vectors, for example P1D11, D11D12, D12D1 of 2 dimensions in new maximum deviation point D11, D12, D21, D22, D31, D32 and formation.Then, whether the decision maximum deviation if 64 below, then for this interval finishes the generative process of 1st hierarchical data at (step S44) below 64, transfers to the generative process of next the 2nd hierarchical data.Under than 64 big situations, newly-generated sub-vector is similarly appended in the 1st hierarchical data as x, y2 amount.Like this, on memory, generate the 1st hierarchical data that generates among Figure 14 A.
In the generative process of the 2nd hierarchical data, as shown in Figure 5, for be judged as maximum deviation in the step S44 of the generative process of the 1st hierarchical data is point below 64, and whether decision is (step S51) below 16 for the absolute value of the maximum deviation of base vector.Be not below 16, promptly than it under the big situation, the variable quantity for the brightness of the sub-vector of benchmark that is evaluated as this point has should be with 6 amount of information that show.On the contrary, under the situation below 16, because being evaluated as the variable quantity for the brightness of the sub-vector of benchmark of this point has and so finish the generative process of the 2nd hierarchical data, transfer to the generative process of the next one the 3rd hierarchical data with 4 or its amount of information with the performance of getting off.
The absolute value of maximum deviation than 16 big situations under, the above-mentioned sub-vector that has generated between each 2 of adjacency that will comprise these points in the step S43 of the 1st hierarchical data generative process is stored in (step S52) in the memory as the 2nd hierarchical data.X, y2 the amount that respectively these sub-vectors is had 6 amount of information as shown in Figure 14 A, arranged these sub-vectors as 1 data continuously.Moreover, in linking each interval separately of 2 of these adjacency, be benchmark with each sub-vector, calculate the deviation of brightness, the point of the maximum deviation of positive and negative is got in decision.
Secondly, at whole maximum deviation point of new decision with constitute among as the sub-vector of the benchmark of these maximum deviation points each 2, respectively similarly generate the sub-vector (step S53) that x, y2 tie up between 2 in adjacency respectively.Whether the absolute value of decision maximum deviation, similarly appends to newly-generated sub-vector in the 2nd hierarchical data as x, y2 amount under than 16 big situations at (step S51) below 16.
Whole maximum deviation point for new decision repeats each step S51～S53, is below 16 up to the absolute value of its maximum deviation.Add successively as the data of the 2nd hierarchical data being judged as, finally on memory, be created on the 2nd hierarchical data shown in Figure 14 A than 16 big sub-vectors.
As shown in Figure 6, similarly carry out the generative process of the 3rd hierarchical data with the generative process of the 2nd hierarchical data.Promptly, for be judged as maximum deviation in the step S51 of the generative process of the 2nd hierarchical data is each point below 16, decision for the absolute value of the maximum deviation of this base vector whether at (step S61) below 4, not below 4, promptly than under its big situation, being evaluated as that variable quantity for the brightness of the sub-vector of benchmark has should be with 4 amount of information that show, on the contrary, under the situation below 4, because being evaluated as that variable quantity for the brightness of the sub-vector of benchmark has should be with 2 or its amount of information with the performance of getting off, finish the generative process of the 3rd hierarchical data, transfer to the generative process of next the 4th hierarchical data.
The absolute value of maximum deviation than 4 big situations under, the above-mentioned sub-vector that has generated between each 2 of adjacency that will comprise these maximum deviation points in the step S53 of the 2nd hierarchical data generative process is stored in (step S62) in the memory as the 3rd hierarchical data.X, y2 the amount that respectively these sub-vectors is had 4 amount of information similarly as shown in Figure 14 A, arranged these sub-vectors continuously as 1 data.Moreover, in linking each interval separately of 2 of these adjacency, be benchmark with each sub-vector, calculate the deviation of brightness, the point of the maximum deviation of positive and negative is got in decision respectively.
Secondly, at whole maximum deviation point of new decision with constitute among as the sub-vector of the benchmark of these maximum deviation points each 2, respectively similarly generate the sub-vector (step S63) that x, y2 tie up between 2 in adjacency respectively.Then, for these new maximum deviation points, whether the absolute value of decision maximum deviation, appends to newly-generated sub-vector in the 3rd hierarchical data as x, y2 amount under than 4 big situations at (step S61) below 4.
For whole maximum deviation point of new decision similarly, repeat each step S61～S63, be below 4 up to the absolute value of its maximum deviation.Add successively as the data of the 3rd hierarchical data being judged as, finally on memory, be created on the 3rd hierarchical data shown in Figure 14 A than 4 big sub-vectors.
As shown in Figure 7, in the generative process of the 4th hierarchical data, be each point below 4 in the step S61 of the generative process of the 3rd hierarchical data, being judged as maximum deviation, decision is 1 or 0 (step S71) for the absolute value of the maximum deviation of this base vector.Be under some situations of 1 or 0, because this point is evaluated as that variable quantity for the brightness of the sub-vector of benchmark has should be with 1 amount of information that shows, so same as Figure 14 A as shown in, the sub-vector that use x, a y2 amount will comprise between 2 of adjacency of this point is stored in (step S72) in the memory as the 4th hierarchical data.
Be not under some situations of 1 or 0 at the absolute value of maximum deviation, the variable quantity that this point is evaluated as for the brightness of the sub-vector of benchmark has than 1 big amount of information.At this moment, in the step S63 of the 3rd hierarchical data generative process, the sub-vector that generates between each 2 of adjacency that will comprise these maximum deviation points calculates the deviation of brightness as benchmark, and the point (step S73) of the maximum deviation of positive and negative is got in decision respectively.Then, at the maximum deviation point of these new decisions with constitute as respectively in 2 of the sub-vector of the benchmark of these maximum deviation points, similarly generate the sub-vector (step S74) of x, y2 dimension between each 2 of adjacency respectively, the absolute value that determines its maximum deviation again is 1 or 0 (step S71).
Repeat these steps S71,73,74, have and be stored in the memory as the 4th hierarchical data with 1 amount of information that shows up to the variable quantity that is evaluated as for the whole maximum deviation points that illustrate like this for the brightness of the sub-vector of benchmark.Thus, the whole point that stays after the generation with the 3rd hierarchical data is transformed to the vector that has with 1 x that shows, y2 amount respectively, generates as the 4th hierarchical data.
Like this,, the brightness of each pixel changed carry out resolution of vectors and handle, generate the 1st～the 4th different hierarchical data of figure place, on memory, arrange in order from the 1st row to the 256th row according to its size for each row of the view data of minimum treat unit.At this moment, in the beginning part of each row configuration about the information of the starting point P1 of original view data and terminal point P2, be the length and the brightness value of transverse direction.In addition, between each hierarchical data, insert the data on its border of expression.
On transverse direction the view data of minimum treat unit being carried out resolution of vectors handles and generates in the said process of the 1st～the 4th hierarchical data, because the view data that is divided into minimum treat unit is independently handled in each row again, so it is less to become the capacity of data of object of 1 processing of having concluded.In addition, if these data are launched in certain zone of memory and handle after constitute once more, then,, can be used in processing next time so opened the zone of these memories owing to do not need original data.Therefore, in the present invention, the memory that effectively utilizes process computer compared with the past, and, can side by side or postpone each row that above-mentioned transverse direction is handled on certain time ground concurrently, so can improve processing speed significantly according to its disposal ability.In addition, even the lower process computer of ability also can be handled view data capacious.
Certainly, the data length of the 1st～the 4th hierarchical data has nothing in common with each other in each row.Therefore, in the present embodiment, for data at all levels, from whole row, select the longest hierarchical data of data length and with it as reference data length, under the data length situation shorter, in Figure 14 B, added the value of putting into " 0 " in the zone of data length deficiency of shade respectively than it.Thus, the view data of minimum treat unit is carried out shaping handles so that the data length of each hierarchical data respectively from the 1st row to the 256th classify as identical, be that each row of longitudinal direction constitutes (step S32) with identical 256 continuous data.
Secondly, will in view data each row of Figure 14 B that has carried out vector quantization on the transverse direction, carry out vector quantization (step S33) successively at longitudinal direction.According to the process shown in Fig. 8, similarly carry out vector quantization of each row with the vector quantization of above-mentioned transverse direction, thus, generate the 1st～the 4th hierarchical data for each row.At first, for 1 row of beginning, with its starting point Q1 (z1=0, y1) and terminal point Q2 (z2=255 y2) be benchmark, generation master vector Q1Q2=[z2-z1, y2-y1] (step S81).
Secondly, in the interval that links above-mentioned starting point and terminal point,, be the deviation that benchmark calculates brightness with the master vector for other pixel, decision get the maximum deviation of positive and negative some V1 (z11, y11), V2 (z12, y12) (step S82).And, between each 2 point of adjacency, Q1 and V1, V1 and V2 and V2 and Q2, similarly generate zy2 respectively and tie up sub-vector Q1V1=[z11-z1, y11-y1], V1V2=[z12-z11, y12-y11] and V2Q2=[z2-z12, y2-y12] (step S83).In addition, each point P1, P2, y1, the y2 of D1, D2, y11, the y12 that uses in this y1 that uses in the each point of Q1, Q2, V1, V2, y2, y11, y12 and vector quantization at transverse direction is different.
With whether relevant above-mentioned process similarly determines the absolute value of the deviation among maximum deviation point V1 and the V2 respectively below 64 with Fig. 4 in the vector of transverse direction, under than 64 big situations, the sub-vector that shows with z, a y2 scale that has 8 amount of information respectively is stored in the memory as the 1st hierarchical data.Then, be benchmark with this sub-vector again, determine the maximum deviation point of new positive and negative, as long as the absolute value of this deviation surpasses 64, just these sub-vectors are appended in the memory, generate the 1st hierarchical data.
Whole maximum deviation point for new decision, if the absolute value of its maximum deviation is below 64, transfer to the generative process of the 2nd hierarchical data equally, carry out vector quantization according to the above-mentioned same process relevant with Fig. 5, generate the 2nd hierarchical data that only constitutes, be stored in the memory by data with amount of information of 6.Secondly, with Fig. 6 relatively with the above-mentioned vector quantization that similarly carries out, generate the 3rd hierarchical data that only constitutes by data with amount of information of 4, it is stored in the memory, afterwards, utilize again with Fig. 7 relatively with above-mentioned same vectorized process, only generate the 4th hierarchical data that constitutes by 1 data, make it be stored in (step S84) in the memory equally.
Even in the vector quantization of this longitudinal direction, also with the vector quantization of transverse direction relatively as described above, for the little data of capacity, in each row, generate each step of the 1st～the 4th hierarchical data independently, therefore, handle original memory block by open data that constitute once more successively and being used in next time, can seek the effective utilization of memory block, and when utilizing each row or the parallel processing that has postponed certain hour realize increasing substantially of processing speed.In addition, even the bigger view data of capacity also can realize the processing of being undertaken by the lower process computer of ability.
Because view data in the enterprising row vector resolution process of longitudinal direction, as shown in Figure 14 B, be divided into starting point and terminal point, the 1st～the 4th hierarchical data in advance, so generate the whole of the 1st～the 4th hierarchical data from starting point and terminal point and the 1st hierarchical data, and generate the 2nd～the 4th hierarchical data from the 2nd hierarchical data, generate the 3rd and the 4th hierarchical data from the 3rd hierarchical data, only generate the 4th hierarchical data from the 4th hierarchical data.In fact, when the application's inventor has carried out the image Compression of present embodiment for several full color test patterns, confirmed that the 1st～the 4th hierarchical data becomes the regional a shown in Figure 15, regional b+c, the such certain ratio of regional d+e+f, regional g+h+i+j substantially.
Secondly, to these data sets a～j, as shown in Figure 16, and by the order of the 1st～the 4th hierarchical data, be that the big order of figure place is carried out shaping abreast and handled, process (step S85) once more on the memory block.Thus, can be from minimum treat unit, be the compressing image data (step S4) that the unit view data of single chrominance channel obtains having desirable matrix structure.
According to the present invention, because the monochrome information of each pixel is carried out vector quantization, and, data are carried out stratification handle and constitute once more, so the monochrome information that does not have the pixel of number of significant digit in fact can be absorbed in 1 vector according to the degree of the number of significant digit of these information.Thereby, only handle by carrying out stratification by this way, just can reduce data volume and the information that comprises in the original file is incurred loss, particularly in the compression of the many high meticulous view data of amount of information, can constitute most effective data structure.In addition, the view data of having carried out 2 n dimensional vector nizations by this way on transverse direction, longitudinal direction is carried out the stratification processing and the processing of formation once more, because by being that benchmark is distinguished data to carry out the stratification processing with certain figure place, so can be described as quantification.
With prior art this view data is compressed processing (step S5).For each hierarchical data of the 1st～the 4th hierarchical data, for example use the arithmetic that has made up known ラ Application レ Application グ ス compression processing and Ha Off マ Application compression processing to compress processing.Thus, the unit view data for single chrominance channel generates packed data of the present invention.According to the present invention, owing to by this way view data has been carried out the stratification processing, thus compared with the past, improved the arithmetic compression effects.Moreover, compress by handle the arithmetic that also carries out except stratification, although be non-loss for view data, can high compression ratio compress as a whole.
For other 2 chrominance channels of the view data that constitutes identical optimization process unit, carry out resolution of vectors similarly respectively and handle, be transformed to the view data of matrix structure, and compress processing, generate packed data of the present invention.
Secondly, in step S6, the image compression data of 3 chrominance channels generating is dividually by this way merged into 1, generate the RGB image compression data of optimization process unit, make it be stored in (step S91) in the memory.If for the generation of 16 whole unit view data end image compression datas, then it is merged into 1, generates the VFZ image file of the present invention (step S92) that has compressed original view data integral body.At this moment, in the beginning part of VFZ image file, insert the heading label of the size recorded and narrated file, data structure etc.
In order to play original coloured image, usually through network with online mode or use the recording medium of CD-R etc., from process computer the VFZ image file is exported to broadcast with computer or work station (step S7).To playing broadcast condition with computer input and output resolution, picture size etc.The VFZ image file is decompressed,, generate the view data of above-mentioned matrix structure according to above-mentioned broadcast condition.From playing the picture of directly view data being exported to the display unit etc. of regulation with computer, or be stored in the memory etc. of server, or transmit with online mode.
Because view data and Figure 16 of being output have above-mentioned matrix structure relatively, so in the moment of having imported the 1st hierarchical data of exporting previously, display image integral body on picture, along with the data below input the 2nd level, show meticulousr image, if in the end import total data, then reproduce original high meticulous coloured image.
In the above-described embodiment, the situation of the view data of RGB system being compressed processing has been described, even but for the coloured image of the form of CMY system and other or have the view data of the such single chrominance channel of gray scale, also can use the present invention equally.In addition, as the practitioner understood, the present invention can increase various changes to the above embodiments, be out of shape and implement in the scope of its technology.
Claims (7)
1. method for compressing image is characterized in that:
Each row for the transverse direction of the bitmap with single or multiple chrominance channel, DID, position and brightness with the transverse direction of each pixel are carried out 2 n dimensional vector nizations to this view data, and be benchmark with the part of line of the Origin And Destination that links above-mentioned each row, the size of utilizing brightness to change quantizes
Each row for the longitudinal direction of above-mentioned view data, position and brightness with longitudinal direction are carried out 2 n dimensional vector nizations to the above-mentioned view data of having carried out vector quantization and quantification on transverse direction, and be benchmark with the part of line of the Origin And Destination that links above-mentioned each row, the size of utilizing brightness to change quantizes, thus, be transformed to the compressing image data of matrix structure of the number of significant digit of the monochrome information that has based on each pixel.
2. the method for compressing image described in claim 1 is characterized in that:
Above-mentioned bitmap, DID are divided into the unit view data of given size,
For the constituent parts view data of having cut apart, carry out the vector quantization and the quantification of above-mentioned transverse direction respectively, and carry out the vector quantization and the quantification of above-mentioned longitudinal direction, be transformed to the unit compressing image data of above-mentioned matrix structure,
Merge above-mentioned constituent parts compressing image data.
3. the method for compressing image described in claim 1 is characterized in that:
Above-mentioned bitmap, DID are divided into each chrominance channel, after the view data of each chrominance channel that will cut apart is transformed to the compressing image data of above-mentioned matrix structure respectively, are merged into 1 compressing image data.
4. the method for compressing image described in claim 2 is characterized in that:
The above-mentioned unit view data of having cut apart is divided into each chrominance channel again, after the unit view data of each chrominance channel that will cut apart is transformed to the unit compressing image data of above-mentioned matrix structure respectively, is merged into 1 unit compressing image data.
5. the method for compressing image described in each of claim 1 to 4 is characterized in that:
Has the process of again compressing image data of above-mentioned matrix structure being carried out the arithmetic compression.
6. the method for compressing image described in each of claim 1 to 4 is characterized in that:
The quantification of above-mentioned transverse direction and longitudinal direction comprises: the part of line with the Origin And Destination that links above-mentioned each row or each row is base vector, calculate in this interval just and/or negative maximum deviation point, in above-mentioned starting point and terminal point and above-mentioned maximum deviation point, with respectively 2 the part of line that links adjacency, be new base vector, the process of calculating this maximum deviation point; And repeat deviation about the base vector of above-mentioned each maximum deviation point till the number of significant digit of forfeiture as desirable view data, utilize the size of the deviation of above-mentioned each maximum deviation point to distinguish each pixel, generate the process of the different many levels data of figure place respectively.
7. the method for compressing image described in claim 6 is characterized in that:
Be under 8 the data conditions at above-mentioned bitmap, DID, above-mentioned many levels data are respectively the 1st～the 4th hierarchical data that is made of 8,6,4,1 data.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP28329699 | 1999-08-27 | ||
| JP283296/99 | 1999-08-27 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN1387722A true CN1387722A (en) | 2002-12-25 |
Family
ID=17663623
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN00814924A Pending CN1387722A (en) | 1999-08-27 | 2000-07-05 | Image compressing method |
Country Status (7)
| Country | Link |
|---|---|
| US (1) | US7031514B1 (en) |
| EP (1) | EP1251686A1 (en) |
| JP (1) | JP3829217B2 (en) |
| KR (1) | KR20020042816A (en) |
| CN (1) | CN1387722A (en) |
| CA (1) | CA2383053A1 (en) |
| WO (1) | WO2001017229A1 (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7639880B2 (en) | 2004-12-10 | 2009-12-29 | Ricoh Company, Ltd. | Compressing a multivalue image with control of memory space requirement |
| CN1787599B (en) * | 2004-12-10 | 2010-05-05 | 株式会社理光 | Image processing apparatus and system, and method for dividing a mutlivalue image for compression |
| CN101123668B (en) * | 2006-08-08 | 2010-12-15 | 上海西门子医疗器械有限公司 | Compression method for CT image |
| CN105938624A (en) * | 2016-04-15 | 2016-09-14 | 张志华 | Fouling monitoring device |
| CN111462258A (en) * | 2020-03-31 | 2020-07-28 | 上海大学 | Texture line image vectorization method and system for manufacturing printed film |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2007178166A (en) * | 2005-12-27 | 2007-07-12 | Rkc Instrument Inc | Method of detecting heater disconnection |
| US8320694B2 (en) * | 2009-03-27 | 2012-11-27 | Sharp Laboratories Of America, Inc. | Surround error diffusion |
| US8274705B2 (en) * | 2009-03-27 | 2012-09-25 | Sharp Laboratories Of America, Inc. | Multi-level surround error diffusion |
| CN107452039B (en) * | 2016-06-01 | 2023-03-31 | 上海东方传媒技术有限公司 | Method and device for compressing RGB color space |
Family Cites Families (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4205341A (en) * | 1978-01-24 | 1980-05-27 | Nippon Telegraph And Telephone Public Corporation | Picture signal coding apparatus |
| JPS57150275A (en) * | 1981-03-11 | 1982-09-17 | Fuji Photo Film Co Ltd | Adaptive quantizer |
| US5072290A (en) * | 1986-09-19 | 1991-12-10 | Canon Kabushiki Kaisha | Color image signal encoding device |
| JPH04579A (en) | 1990-04-17 | 1992-01-06 | Matsushita Electric Ind Co Ltd | Method for extracting feature point of graphic |
| US5828784A (en) * | 1992-07-13 | 1998-10-27 | Hitachi Denshi Kabushiki Kaisha | Data coding method and apparatus using a plurality of blocks of different length |
| JPH08194816A (en) | 1995-01-17 | 1996-07-30 | Pesupu Purosu Kk | Segment approximation method and its system |
| JP3448601B2 (en) | 1996-04-30 | 2003-09-22 | インターネットナンバー株式会社 | Multi-tone image data compression / decompression method |
| JPH1032719A (en) | 1996-07-18 | 1998-02-03 | Benetsuse Corp:Kk | Picture compression method and device therefor |
| JP3957915B2 (en) * | 1999-03-08 | 2007-08-15 | パイオニア株式会社 | Fade detection device and information encoding device |
| US6868190B1 (en) * | 2000-10-19 | 2005-03-15 | Eastman Kodak Company | Methods for automatically and semi-automatically transforming digital image data to provide a desired image look |
| WO2002043401A1 (en) * | 2000-11-22 | 2002-05-30 | Koninklijke Philips Electronics N.V. | Video signal processing |
| FI112424B (en) * | 2001-10-30 | 2003-11-28 | Oplayo Oy | Coding procedure and arrangement |
-
2000
- 2000-07-05 CN CN00814924A patent/CN1387722A/en active Pending
- 2000-07-05 CA CA002383053A patent/CA2383053A1/en not_active Abandoned
- 2000-07-05 EP EP00944269A patent/EP1251686A1/en not_active Withdrawn
- 2000-07-05 KR KR1020027002551A patent/KR20020042816A/en not_active Application Discontinuation
- 2000-07-05 US US10/069,227 patent/US7031514B1/en not_active Expired - Fee Related
- 2000-07-05 JP JP2001521051A patent/JP3829217B2/en not_active Expired - Fee Related
- 2000-07-05 WO PCT/JP2000/004472 patent/WO2001017229A1/en not_active Application Discontinuation
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7639880B2 (en) | 2004-12-10 | 2009-12-29 | Ricoh Company, Ltd. | Compressing a multivalue image with control of memory space requirement |
| CN1787599B (en) * | 2004-12-10 | 2010-05-05 | 株式会社理光 | Image processing apparatus and system, and method for dividing a mutlivalue image for compression |
| CN101123668B (en) * | 2006-08-08 | 2010-12-15 | 上海西门子医疗器械有限公司 | Compression method for CT image |
| CN105938624A (en) * | 2016-04-15 | 2016-09-14 | 张志华 | Fouling monitoring device |
| CN111462258A (en) * | 2020-03-31 | 2020-07-28 | 上海大学 | Texture line image vectorization method and system for manufacturing printed film |
| CN111462258B (en) * | 2020-03-31 | 2023-09-15 | 上海大学 | Texture line image vectorization method and system for manufacturing printing film |
Also Published As
| Publication number | Publication date |
|---|---|
| EP1251686A1 (en) | 2002-10-23 |
| KR20020042816A (en) | 2002-06-07 |
| CA2383053A1 (en) | 2001-03-08 |
| US7031514B1 (en) | 2006-04-18 |
| JP3829217B2 (en) | 2006-10-04 |
| WO2001017229A1 (en) | 2001-03-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1093669C (en) | A histogram equalization apparatus for contrast enhancement of moving image and method therefor | |
| JP3978478B2 (en) | Apparatus and method for performing fixed-speed block-unit image compression with estimated pixel values | |
| CN1106766C (en) | Method and apparatus for generating chrominance shape information of video object plane in video signal | |
| CN1012302B (en) | Color image dispaly system | |
| JP4100782B2 (en) | Error diffusion value processing method | |
| CN104012097A (en) | Method and apparatus for compressing and decompressing data | |
| WO1994000949B1 (en) | Video compression and decompression using block selection and subdivision | |
| CN1389840A (en) | Method and apparatus for treating video picture data displayed on display device | |
| CN1070530A (en) | The device of splitting video signal between two channels | |
| CN1387722A (en) | Image compressing method | |
| CN100474936C (en) | Apparatus and method for transforming three color signals into multi-color signal | |
| CN112118449B (en) | Method and device for compressing and decompressing image | |
| CN102473314A (en) | Vector embedded graphics coding | |
| CN1770874A (en) | Apparatus and method for controlling hue and luminance in image display device | |
| CN101064816A (en) | Image processing apparatus and method for preventing degradation of image quality when bit format of image is converted | |
| Makhinya et al. | Fast compositing for cluster-parallel rendering | |
| JPH06164950A (en) | Picture coding method and picture coding/decoding method | |
| Rui et al. | On the initialization and training methods for Kohonen self-organizing feature maps in color image quantization | |
| CN1254772C (en) | System and method for optimizing semi-tone printer performance | |
| US6970176B1 (en) | Video processing in PC uses statistically tuned color cube | |
| JP2003309727A5 (en) | ||
| CN1073262C (en) | Parallel memory device for image processing utilizing linear transformation | |
| CN110689590B (en) | Compression coding method and device for map and mobile terminal equipment | |
| JP2521780B2 (en) | Image information processing device | |
| CN1273948C (en) | Method for changing image definition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |